Automatic Depression Prediction using Screen Lock/Unlock Data on the Smartphone

Computer passwords as a timely booster for writing-based psychological interventions

“…The current study had several limitations. First, we used retrospective reports of password usage frequency, which might have contributed to self-recall bias; it would be ideal to have an objectively measured variable, such as screen lock/unlock data ( Kim et al, 2021 ), or to use daily diary methods ( Gunthert and Wenze, 2012 ) to sample self-reports of daily password usage on different days during the study period. Second, password usage frequency was self-reported rather than manipulated, constraining our ability to draw causal inferences; it is possible, for example, that high levels of psychological well-being led to higher productivity and therefore an increased use of self-affirming passwords to unlock the computers.…”

Construction method of college students' depression knowledge map based on education big data

“…However, many machine learning models that predict human characteristics do not fully consider the role of pre training language models and context embedding. Kim et al [6] took a person's degree of depression as a case study and conducted empirical analysis to determine which ready-made language model, single-layer and multi-layer combination seems to be the most promising when applied to human level NLP tasks. Dai et al [7] developed a new two-stage feature selection algorithm, which is based on highdimensional (more than 30000) features constructed through context aware analysis of DAIC-WOZ dataset (including audio, video and semantic features), compared the prediction performance with seven reference models, and analyzed the preferred topics and functional categories related to reserved functions.…”